]> granicus.if.org Git - postgresql/log
postgresql
20 years agoThe following bug has been logged online:
Bruce Momjian [Fri, 20 Feb 2004 20:42:29 +0000 (20:42 +0000)]
The following bug has been logged online:

Bug reference:      1081
Logged by:          Aarjav Trivedi

Email address:      aarjav@cc.gatech.edu

PostgreSQL version: 7.4

Operating system:   Linux

Description:        Spelling error in tsearch2.sql leading to problems
with
tsearch

Details:

On line 620 of tsearch2.sql which is required to install and run
TSEARCH,

REATE FUNCTION tsstat_in(cstring)

should be

CREATE FUNCTION tsstat_in(cstring)

because of this error, TSEARCH fails to work as specified,

20 years agoAdd:
Bruce Momjian [Fri, 20 Feb 2004 20:37:11 +0000 (20:37 +0000)]
Add:

> * Have psql show more information about sequences

20 years agoAdd:
Bruce Momjian [Fri, 20 Feb 2004 18:45:35 +0000 (18:45 +0000)]
Add:

> * Allow external interfaces to extend the GUC variable set

20 years agoAdd to PL/java description.
Bruce Momjian [Fri, 20 Feb 2004 18:31:03 +0000 (18:31 +0000)]
Add to PL/java description.

20 years agoRe-implement psql's input scanning to use a flex-generated lexer, as per
Tom Lane [Thu, 19 Feb 2004 19:40:09 +0000 (19:40 +0000)]
Re-implement psql's input scanning to use a flex-generated lexer, as per
recent discussion.  The lexer is used for both SQL command text and
backslash commands.  The purpose of this change is to make it easier to
track the behavior of the backend's SQL lexer --- essentially identical
flex rules are now used by psql.  Also, this cleans up a lot of very
squirrelly code in mainloop.c and command.c.  The flex code is somewhat
bulkier than the removed code, but should be lots easier to maintain.

20 years agoCosmetic changes (mostly whitespace) to make it easier to diff the
Tom Lane [Thu, 19 Feb 2004 19:11:30 +0000 (19:11 +0000)]
Cosmetic changes (mostly whitespace) to make it easier to diff the
backend lexer against psql's.

20 years agoUpdate:
Bruce Momjian [Wed, 18 Feb 2004 21:42:22 +0000 (21:42 +0000)]
Update:

< * Add GUC variable to prevent waiting on locks
> * Add NO WAIT option to various SQL commands

20 years agoHere is a patch that implements setitimer() on win32. With this patch
Bruce Momjian [Wed, 18 Feb 2004 16:25:12 +0000 (16:25 +0000)]
Here is a patch that implements setitimer() on win32. With this patch
applied, deadlock detection and statement_timeout now works.

The file timer.c goes into src/backend/port/win32/.

The patch also removes two lines of "printf debugging" accidentally left
in pqsignal.h, in the console control handler.

Magnus Hagander

20 years agoAdded missing rule for DOUBLE variables.
Michael Meskes [Wed, 18 Feb 2004 08:42:02 +0000 (08:42 +0000)]
Added missing rule for DOUBLE variables.

20 years agoImprove the consistency of the error message emitted when rejecting
Neil Conway [Wed, 18 Feb 2004 00:01:34 +0000 (00:01 +0000)]
Improve the consistency of the error message emitted when rejecting
invalid input to the oid type. Also, remove some long-unused code
from adt/numutils.c

20 years agoRemove a caveat from the "backup" documentation: pg_dump now does a
Neil Conway [Tue, 17 Feb 2004 23:56:07 +0000 (23:56 +0000)]
Remove a caveat from the "backup" documentation: pg_dump now does a
better job of handling dependencies between database objects.

20 years agoDone:
Bruce Momjian [Tue, 17 Feb 2004 15:46:44 +0000 (15:46 +0000)]
Done:
> * -Change CVS ID to PostgreSQL

20 years agoAdd:
Bruce Momjian [Tue, 17 Feb 2004 15:37:48 +0000 (15:37 +0000)]
Add:

>  o Allow the schema of objects to be changed

20 years agoAdd a few more <xref> tags to the SGML docs, where appropriate. Original
Neil Conway [Tue, 17 Feb 2004 09:07:16 +0000 (09:07 +0000)]
Add a few more <xref> tags to the SGML docs, where appropriate. Original
patch from Michael Glaesemann, additional changes by Neil Conway.

20 years agoMinor editorializing on cost-based vacuum description.
Tom Lane [Tue, 17 Feb 2004 07:36:47 +0000 (07:36 +0000)]
Minor editorializing on cost-based vacuum description.

20 years agoSignificant improvements to the documentation for the new cost-based
Neil Conway [Tue, 17 Feb 2004 06:28:05 +0000 (06:28 +0000)]
Significant improvements to the documentation for the new cost-based
vacuum delay feature, including updating the docs for Tom's recent
improvements. There is still more work to be done here: for example,
adding some more information on the practical use of cost-based
vacuum delay to the "maintenance" section would probably be a good
idea.

20 years agoFix a minor SGML markup infelicity in recently-applied log_disconnections
Neil Conway [Tue, 17 Feb 2004 05:45:17 +0000 (05:45 +0000)]
Fix a minor SGML markup infelicity in recently-applied log_disconnections
patch.

20 years agoRename function log_session_end to log_disconnections.
Bruce Momjian [Tue, 17 Feb 2004 04:09:26 +0000 (04:09 +0000)]
Rename function log_session_end to log_disconnections.

20 years agoFix prototype for on_proc_exit in log_disconnections patch.
Bruce Momjian [Tue, 17 Feb 2004 04:06:26 +0000 (04:06 +0000)]
Fix prototype for on_proc_exit in log_disconnections patch.

20 years agoThis patch brings up to date what I did last year (now unfortunately
Bruce Momjian [Tue, 17 Feb 2004 03:54:57 +0000 (03:54 +0000)]
This patch brings up to date what I did last year (now unfortunately
bitrotted) to allow the logging of the end of a session, enabled by
the config setting "log_disconnections".

Andrew Dunstan

20 years agoHere is an updated version of the win32 readdir patch.
Bruce Momjian [Tue, 17 Feb 2004 03:45:17 +0000 (03:45 +0000)]
Here is an updated version of the win32 readdir patch.

1) Now puts in exactly the same change as the current-cvs mingw code
does. (see
http://cvs.sourceforge.net/viewcvs.py/mingw/runtime/mingwex/dirent.c?r1=
1.3&r2=1.4, second part of the patch).

2) Updates both xlog.c and slru.c in backend/access/transam/

3) Also updates pg_resetxlog, which also uses readdir() and checks the
errno value after the loop.

Magnus Hagander

20 years agoUnder Win32, stat() returns an st_ino field, but it has no meaning (on
Bruce Momjian [Tue, 17 Feb 2004 03:35:57 +0000 (03:35 +0000)]
Under Win32, stat() returns an st_ino field, but it has no meaning (on
Win2K, and possibly all Win32 variants, it is always 0). This causes a
number of problems in the dfmgr.c logic, which basically all revolve
around the fact that *any* two files will appear to have the same inode.

Claudio Natoli

20 years agoPlease apply this patch to contrib/dbmirror
Bruce Momjian [Tue, 17 Feb 2004 03:34:35 +0000 (03:34 +0000)]
Please apply this patch to contrib/dbmirror
In incorperates changes from myself and a number of contributors.

This update to dbmirror provides:

-replication of sequence operations via setval/nextval
-DBMirror.pl support for logging to syslog
-changed the names of the tables to dbmirror_*  (no quotes required)
-Support for writitng SQL statements to files instead of directly to
 a slave database
-More options for DBMirror.pl in the config files.

Steven Singer

20 years agoMinor SGML markup improvements.
Neil Conway [Tue, 17 Feb 2004 02:53:03 +0000 (02:53 +0000)]
Minor SGML markup improvements.

20 years agoMake use of statistics on index expressions. There are still some
Tom Lane [Tue, 17 Feb 2004 00:52:53 +0000 (00:52 +0000)]
Make use of statistics on index expressions.  There are still some
corner cases that could stand improvement, but it does all the basic
stuff.  A byproduct is that the selectivity routines are no longer
constrained to working on simple Vars; we might in future be able to
improve the behavior for subexpressions that don't match indexes.

20 years agoChange may -> might, and remove an extra tab.
Bruce Momjian [Mon, 16 Feb 2004 23:07:57 +0000 (23:07 +0000)]
Change may -> might, and remove an extra tab.

20 years ago- Cleaned up parser a little bit. It does not make sense to allow a
Michael Meskes [Mon, 16 Feb 2004 07:41:54 +0000 (07:41 +0000)]
- Cleaned up parser a little bit. It does not make sense to allow a
  typename to be typedef'ed that cannot be parsed as variable type.
- Allowed some SQL keywords to be used as C variable names.

20 years agoFirst steps towards statistics on expressional (nee functional) indexes.
Tom Lane [Sun, 15 Feb 2004 21:01:39 +0000 (21:01 +0000)]
First steps towards statistics on expressional (nee functional) indexes.
This commit teaches ANALYZE to store such stats in pg_statistic, but
nothing is done yet about teaching the planner to use 'em.
Also, repair longstanding oversight in separate ANALYZE command: it
updated the pg_class.relpages and reltuples counts for the table proper,
but not for indexes.

20 years agoNew translation
Peter Eisentraut [Sun, 15 Feb 2004 19:38:29 +0000 (19:38 +0000)]
New translation

20 years ago- Allowed some C keywords to be used as SQL column names.
Michael Meskes [Sun, 15 Feb 2004 15:38:20 +0000 (15:38 +0000)]
- Allowed some C keywords to be used as SQL column names.

20 years ago - Added missing braces to array parsing.
Michael Meskes [Sun, 15 Feb 2004 13:48:54 +0000 (13:48 +0000)]
    - Added missing braces to array parsing.
- Set ecpg version to 3.1.1.
- Removed that old debugging output that I forgot the last time.

20 years agoMinor documentation tweak.
Neil Conway [Sun, 15 Feb 2004 06:27:37 +0000 (06:27 +0000)]
Minor documentation tweak.

20 years agoCreate crosstype comparison operators for date vs. timestamp and date
Tom Lane [Sat, 14 Feb 2004 20:16:18 +0000 (20:16 +0000)]
Create crosstype comparison operators for date vs. timestamp and date
vs. timestamptz.  This allows use of indexes for expressions like
  datecol >= date 'today' - interval '1 month'
which were formerly not indexable without casting the righthand side
down from timestamp to date.

20 years agoRepair optimization bug I introduced in a moment of brain fade back in
Tom Lane [Fri, 13 Feb 2004 22:26:30 +0000 (22:26 +0000)]
Repair optimization bug I introduced in a moment of brain fade back in
Nov 2002: when constant-expression simplification removes all the
aggregate function calls from a query, that doesn't mean we can act as
though there never were any aggregates.  Per bug report from Gabor Szucs.

20 years agoDocumentation for vacuum_cost config options.
Jan Wieck [Fri, 13 Feb 2004 12:25:09 +0000 (12:25 +0000)]
Documentation for vacuum_cost config options.

Jan

20 years agoAfter further thought about support for gathering stats on functional
Tom Lane [Fri, 13 Feb 2004 06:39:49 +0000 (06:39 +0000)]
After further thought about support for gathering stats on functional
indexes, it seems like we ought to put another layer of indirection
between the compute_stats functions and the actual data storage.  This
would allow us to compute the values on-the-fly, for example.

20 years agoMention psql variable names are case-sensitive.
Bruce Momjian [Fri, 13 Feb 2004 05:10:02 +0000 (05:10 +0000)]
Mention psql variable names are case-sensitive.

20 years agoFix typo:
Bruce Momjian [Fri, 13 Feb 2004 04:47:13 +0000 (04:47 +0000)]
Fix typo:

< * Allow cross-db queries with transaction sematics
> * Allow cross-db queries with transaction semantics

20 years agoTom Lane wrote:
Bruce Momjian [Fri, 13 Feb 2004 04:40:06 +0000 (04:40 +0000)]
Tom Lane wrote:
> momjian@svr1.postgresql.org (Bruce Momjian) writes:
>>      someone asked me about the FK deadlock fix, mentioned in the 7.3.3
>>      release notes as 3rd change:
>>      http://www.postgresql.org/docs/current/static/release-7-3-3.html
>>      Actually, that fix was available with 7.4, not 7.3. Don't know if we can
>>      retroactively change the release-notes though.
>
> This is completely erroneous, please undo it.
>
> 2003-05-21 14:14  tgl
>
>       * src/: backend/utils/adt/ri_triggers.c,
>       test/regress/expected/foreign_key.out (REL7_3_STABLE): Back-patch
>       Jan's fix to avoid primary key lookup (and lock) if foreign key
>       does not change on UPDATE.

Oh ... didn't know that you did a backpatch. Sorry

Jan

20 years agoAdd:
Bruce Momjian [Fri, 13 Feb 2004 04:36:02 +0000 (04:36 +0000)]
Add:

> * Use nested transactions to prevent syntax errors from aborting a transaction

20 years agoAdd:
Bruce Momjian [Fri, 13 Feb 2004 02:30:53 +0000 (02:30 +0000)]
Add:

> * Allow temporary views on non-temporary tables

20 years agoTry to make 'cross-database references are not implemented' errors a
Tom Lane [Fri, 13 Feb 2004 01:08:20 +0000 (01:08 +0000)]
Try to make 'cross-database references are not implemented' errors a
bit more helpful by identifying the specific qualified name being
complained of.

20 years agoAdd hooks for type-specific calculation of ANALYZE statistics. Idea and
Tom Lane [Thu, 12 Feb 2004 23:41:04 +0000 (23:41 +0000)]
Add hooks for type-specific calculation of ANALYZE statistics.  Idea and
coding by Mark Cave-Ayland, some kibitzing by Tom Lane.  initdb forced
due to new column in pg_type.

20 years agoAdd from Joe Conway:
Bruce Momjian [Thu, 12 Feb 2004 23:27:52 +0000 (23:27 +0000)]
Add from Joe Conway:

>  o Modify array literal representation to handle array index lower bound
>    of other than one
>
183d185
<

20 years agoAdd CVS ID tags to port/win32/files.
Bruce Momjian [Thu, 12 Feb 2004 20:37:34 +0000 (20:37 +0000)]
Add CVS ID tags to port/win32/files.

20 years agoJan:
Bruce Momjian [Thu, 12 Feb 2004 20:17:34 +0000 (20:17 +0000)]
Jan:

someone asked me about the FK deadlock fix, mentioned in the 7.3.3
release notes as 3rd change:

     http://www.postgresql.org/docs/current/static/release-7-3-3.html

Actually, that fix was available with 7.4, not 7.3. Don't know if we can
retroactively change the release-notes though.

20 years agoAvoid delaying postmaster shutdown by up to 10 seconds on platforms
Tom Lane [Thu, 12 Feb 2004 20:07:26 +0000 (20:07 +0000)]
Avoid delaying postmaster shutdown by up to 10 seconds on platforms
where signals do not terminate sleep() delays.

20 years agoImprove \? wording, remove 'internal'.
Bruce Momjian [Thu, 12 Feb 2004 19:58:16 +0000 (19:58 +0000)]
Improve \? wording, remove 'internal'.

20 years agoTranslation updates
Peter Eisentraut [Thu, 12 Feb 2004 19:19:32 +0000 (19:19 +0000)]
Translation updates

20 years agoAdd bitmap discussion to performance TODO.detail.
Bruce Momjian [Thu, 12 Feb 2004 18:13:29 +0000 (18:13 +0000)]
Add bitmap discussion to performance TODO.detail.

20 years agoRemove TODO.detail files that contained useless or very old information.
Bruce Momjian [Thu, 12 Feb 2004 18:11:54 +0000 (18:11 +0000)]
Remove TODO.detail files that contained useless or very old information.
Update TODO accordingly.

20 years agoDone:
Bruce Momjian [Thu, 12 Feb 2004 18:11:16 +0000 (18:11 +0000)]
Done:

<
<
< Reporting
< =========
<
< * Show location of syntax error in query [yacc]

20 years agoFile not needed:
Bruce Momjian [Thu, 12 Feb 2004 18:09:43 +0000 (18:09 +0000)]
File not needed:

.< * Improve speed with indexes (perhaps recreate index instead) [vacuum]
> * Improve speed with indexes (perhaps recreate index instead)
369c369
<   lock and truncate table [vacuum]
>   lock and truncate table
371c371
<   rather than in /contrib [vacuum]
>   rather than in /contrib

20 years agoMostly done:
Bruce Momjian [Thu, 12 Feb 2004 18:05:51 +0000 (18:05 +0000)]
Mostly done:

< * Allow better handling of numeric constants, type conversion [typeconv]

20 years agoFile not needed:
Bruce Momjian [Thu, 12 Feb 2004 18:01:20 +0000 (18:01 +0000)]
File not needed:

< * Add replication of distributed databases [replication]
> * Add replication of distributed databases

20 years agoRemove:
Bruce Momjian [Thu, 12 Feb 2004 17:59:50 +0000 (17:59 +0000)]
Remove:

<  o http://gborg.postgresql.org/project/pgreplication/projdisplay.php

20 years agoMerge files:
Bruce Momjian [Thu, 12 Feb 2004 17:51:03 +0000 (17:51 +0000)]
Merge files:

< * Allow persistent backends [persistent]
> * Allow persistent backends [pool]
397c397
<   connection pooling
>   connection pooling [pool]

20 years agoNot needed:
Bruce Momjian [Thu, 12 Feb 2004 17:38:50 +0000 (17:38 +0000)]
Not needed:

<   float4, numeric/decimal too [optimizer]
>   float4, numeric/decimal too
422c422
< * Missing optimizer selectivities for date, r-tree, etc. [optimizer]
> * Missing optimizer selectivities for date, r-tree, etc

20 years agoNot needed:
Bruce Momjian [Thu, 12 Feb 2004 17:32:38 +0000 (17:32 +0000)]
Not needed:

< * Delay fsync() when other backends are about to commit too [fsync]
> * Delay fsync() when other backends are about to commit too

20 years agoNot needed:
Bruce Momjian [Thu, 12 Feb 2004 17:31:09 +0000 (17:31 +0000)]
Not needed:

< * Add MATCH PARTIAL referential integrity [foreign]
> * Add MATCH PARTIAL referential integrity

20 years agoRemove crossdb content.
Bruce Momjian [Thu, 12 Feb 2004 17:23:30 +0000 (17:23 +0000)]
Remove crossdb content.

20 years agoAdd, dblink doesn't do this:
Bruce Momjian [Thu, 12 Feb 2004 17:22:39 +0000 (17:22 +0000)]
Add, dblink doesn't do this:

> * Allow cross-db queries with transaction sematics

20 years agoDone, mostly by dblink, I think:
Bruce Momjian [Thu, 12 Feb 2004 17:21:41 +0000 (17:21 +0000)]
Done, mostly by dblink, I think:

< * Allow queries across multiple databases [crossdb]

20 years agoFix typo:
Bruce Momjian [Thu, 12 Feb 2004 16:38:04 +0000 (16:38 +0000)]
Fix typo:

On Mon, 2004-01-26 at 21:28, Peter Eisentraut wrote:
> Christophe Combelles wrote:
> > At the bottom of this doc file :
> > file:///usr/share/doc/postgresql-doc/html/tutorial-createdb.html
> > "and it also happens that that user always has permission"
> >                       ---- x2
> The first "that" serves as a conjuction, the second one serves as an
> article.  Looks correct to me.

A better workaround for the sentence would be something like:

        "and it also happens that the user always has permission"

Looks easier to read, I think.

Enver ALTIN (a.k.a. skyblue)

20 years agoAdd:
Bruce Momjian [Thu, 12 Feb 2004 16:05:15 +0000 (16:05 +0000)]
Add:
<
> * Allow substring/replace() to get/set bit values

20 years agoFixed bug where FlushRelationBuffers() did call StrategyInvalidateBuffer()
Jan Wieck [Thu, 12 Feb 2004 15:06:56 +0000 (15:06 +0000)]
Fixed bug where FlushRelationBuffers() did call StrategyInvalidateBuffer()
for already empty buffers because their buffer tag was not cleard out
when the buffers have been invalidated before.

Also removed the misnamed BM_FREE bufhdr flag and replaced the checks,
which effectively ask if the buffer is unpinned, with checks against the
refcount field.

Jan

20 years agoRemove URL for specific pljava implementation:
Bruce Momjian [Thu, 12 Feb 2004 14:41:13 +0000 (14:41 +0000)]
Remove URL for specific pljava implementation:

<  o Allow Java server-side programming, http://pljava.sourceforge.net
<    [java]
>  o Allow Java server-side programming [java]

20 years agoCheck only live tuples for OID validity, to avoid issues after ALTER
Tom Lane [Thu, 12 Feb 2004 05:39:55 +0000 (05:39 +0000)]
Check only live tuples for OID validity, to avoid issues after ALTER
TABLE.  Per recent discussion.

20 years agoAdded hints about the reason, why the command string in
Jan Wieck [Thu, 12 Feb 2004 01:44:22 +0000 (01:44 +0000)]
Added hints about the reason, why the command string in
the view pg_stat_activity is missing, as per Bruces suggestion.

Jan

20 years agoCommit the reasonably uncontroversial parts of J.R. Nield's PITR patch, to
Tom Lane [Wed, 11 Feb 2004 22:55:26 +0000 (22:55 +0000)]
Commit the reasonably uncontroversial parts of J.R. Nield's PITR patch, to
wit: Add a header record to each WAL segment file so that it can be reliably
identified.  Avoid splitting WAL records across segment files (this is not
strictly necessary, but makes it simpler to incorporate the header records).
Make WAL entries for file creation, deletion, and truncation (as foreseen but
never implemented by Vadim).  Also, add support for making XLOG_SEG_SIZE
configurable at compile time, similarly to BLCKSZ.  Fix a couple bugs I
introduced in WAL replay during recent smgr API changes.  initdb is forced
due to changes in pg_control contents.

20 years agoRepair some problems in bgwriter start/stop logic. In particular, don't
Tom Lane [Wed, 11 Feb 2004 22:25:02 +0000 (22:25 +0000)]
Repair some problems in bgwriter start/stop logic.  In particular, don't
allow the bgwriter to start before the startup subprocess has finished
... it tends to crash otherwise.  (The same problem may have existed for
the checkpointer, I'm not entirely sure.)  Remove some code that was
redundant because the bgwriter is handled as a member of the backend list.

20 years agoAdd:
Bruce Momjian [Wed, 11 Feb 2004 22:04:35 +0000 (22:04 +0000)]
Add:

>  o Disallow dropping of an inherited constraint

20 years agoCheck and set thread-safe functions separately, rather than as a single
Bruce Momjian [Wed, 11 Feb 2004 21:44:06 +0000 (21:44 +0000)]
Check and set thread-safe functions separately, rather than as a single
variable.

Remove thread locking for non-thread-safe functions, instead throw a
compile error.

Platforms will have to re-run tools/thread to record their thread
safety.

20 years agoProperly set NEED_REENTRANT_FUNCS for threaded libpq/ecpg.
Bruce Momjian [Wed, 11 Feb 2004 17:29:01 +0000 (17:29 +0000)]
Properly set NEED_REENTRANT_FUNCS for threaded libpq/ecpg.

Without this patch, no thread locking or *_r functions were being used.

20 years agoDoes not apply anymore:
Bruce Momjian [Wed, 11 Feb 2004 17:04:04 +0000 (17:04 +0000)]
Does not apply anymore:

< * Make blind writes go through the file descriptor cache

20 years agoNo need to link libpq into these programs ...
Tom Lane [Tue, 10 Feb 2004 23:24:13 +0000 (23:24 +0000)]
No need to link libpq into these programs ...

20 years agoSGML markup improvements.
Neil Conway [Tue, 10 Feb 2004 20:51:39 +0000 (20:51 +0000)]
SGML markup improvements.

20 years agoAdd configure support for determining UINT64_FORMAT, the appropriate
Tom Lane [Tue, 10 Feb 2004 19:55:45 +0000 (19:55 +0000)]
Add configure support for determining UINT64_FORMAT, the appropriate
snprintf format for uint64 items.

20 years agoMore clearly describe why PQinSend returns false for non-threaded libpq.
Bruce Momjian [Tue, 10 Feb 2004 15:21:24 +0000 (15:21 +0000)]
More clearly describe why PQinSend returns false for non-threaded libpq.

20 years agoAdd URL for most current version.
Bruce Momjian [Tue, 10 Feb 2004 15:16:36 +0000 (15:16 +0000)]
Add URL for most current version.

20 years agoRepair missing inclusions of -lintl for shared libraries.
Tom Lane [Tue, 10 Feb 2004 07:26:25 +0000 (07:26 +0000)]
Repair missing inclusions of -lintl for shared libraries.

20 years agoLooks like we need <unistd.h> for select() on Darwin.
Tom Lane [Tue, 10 Feb 2004 04:23:03 +0000 (04:23 +0000)]
Looks like we need <unistd.h> for select() on Darwin.

20 years agoCentralize implementation of delay code by creating a pg_usleep()
Tom Lane [Tue, 10 Feb 2004 03:42:45 +0000 (03:42 +0000)]
Centralize implementation of delay code by creating a pg_usleep()
subroutine in src/port/pgsleep.c.  Remove platform dependencies from
miscadmin.h and put them in port.h where they belong.  Extend recent
vacuum cost-based-delay patch to apply to VACUUM FULL, ANALYZE, and
non-btree index vacuuming.

By the way, where is the documentation for the cost-based-delay patch?

20 years agoRestructure smgr API as per recent proposal. smgr no longer depends on
Tom Lane [Tue, 10 Feb 2004 01:55:27 +0000 (01:55 +0000)]
Restructure smgr API as per recent proposal.  smgr no longer depends on
the relcache, and so the notion of 'blind write' is gone.  This should
improve efficiency in bgwriter and background checkpoint processes.
Internal restructuring in md.c to remove the not-very-useful array of
MdfdVec objects --- might as well just use pointers.
Also remove the long-dead 'persistent main memory' storage manager (mm.c),
since it seems quite unlikely to ever get resurrected.

20 years agoWin32 signals cleanup. Patch by Magnus Hagander, with input from Claudio
Neil Conway [Sun, 8 Feb 2004 22:28:57 +0000 (22:28 +0000)]
Win32 signals cleanup. Patch by Magnus Hagander, with input from Claudio
Natoli and Bruce Momjian (and some cosmetic fixes from Neil Conway).
Changes:

    - remove duplicate signal definitions from pqsignal.h

    - replace pqkill() with kill() and redefine kill() in Win32

    - use ereport() in place of fprintf() in some error handling in
      pqsignal.c

    - export pg_queue_signal() and make use of it where necessary

    - add a console control handler for Ctrl-C and similar handling
      on Win32

    - do WaitForSingleObjectEx() in CHECK_FOR_INTERRUPTS() on Win32;
      query cancelling should now work on Win32

    - various other fixes and cleanups

20 years agoAdd mention of the developer's FAQ.
Bruce Momjian [Sun, 8 Feb 2004 03:16:31 +0000 (03:16 +0000)]
Add mention of the developer's FAQ.

20 years agoAdd mention of PostgreSQL as a community project.
Bruce Momjian [Sun, 8 Feb 2004 03:09:12 +0000 (03:09 +0000)]
Add mention of PostgreSQL as a community project.

20 years agoAdd PITR web page:
Bruce Momjian [Sun, 8 Feb 2004 01:17:09 +0000 (01:17 +0000)]
Add PITR web page:

< * Point-in-time data recovery using backup and write-ahead log
< * Create native Win32 port, http://momjian.postgresql.org/main/writings/pgsql/win32.html
> * Point-in-time data recovery using backup and write-ahead log,
>   http://momjian.postgresql.org/main/writings/pgsql/project/pitr.html
> * Create native Win32 port, http://momjian.postgresql.org/main/writings/pgsql/project/win32.html

20 years agoAdd missing dependency, per Wade Klaver.
Tom Lane [Sat, 7 Feb 2004 07:20:12 +0000 (07:20 +0000)]
Add missing dependency, per Wade Klaver.

20 years agoCost based vacuum delay feature.
Jan Wieck [Fri, 6 Feb 2004 19:36:18 +0000 (19:36 +0000)]
Cost based vacuum delay feature.

Jan

20 years agoDocumentation for generate_series() functions committed a few days ago.
Joe Conway [Thu, 5 Feb 2004 22:54:36 +0000 (22:54 +0000)]
Documentation for generate_series() functions committed a few days ago.

20 years agoBacking out the background writer sync() option.
Jan Wieck [Wed, 4 Feb 2004 01:24:53 +0000 (01:24 +0000)]
Backing out the background writer sync() option.

Jan

20 years agoUse memmove() rather than memcpy() in set_var_from_var(). If this function
Neil Conway [Wed, 4 Feb 2004 01:11:47 +0000 (01:11 +0000)]
Use memmove() rather than memcpy() in set_var_from_var(). If this function
is asked to assign a variable to itself, it will result in doing a
memcpy() on an entirely-overlapping memory range, which results in
undefined behavior according to ANSI C. That said, it is unlikely to
actually do anything bad on any sane libc, but this keeps valgrind quiet.

20 years agopwd
Tom Lane [Tue, 3 Feb 2004 17:52:55 +0000 (17:52 +0000)]
pwd

20 years agoRename SortMem and VacuumMem to work_mem and maintenance_work_mem.
Tom Lane [Tue, 3 Feb 2004 17:34:04 +0000 (17:34 +0000)]
Rename SortMem and VacuumMem to work_mem and maintenance_work_mem.
Make btree index creation and initial validation of foreign-key constraints
use maintenance_work_mem rather than work_mem as their memory limit.
Add some code to guc.c to allow these variables to be referenced by their
old names in SHOW and SET commands, for backwards compatibility.

20 years agoAdd:
Bruce Momjian [Tue, 3 Feb 2004 16:53:56 +0000 (16:53 +0000)]
Add:

> * Make pg_restore continue after errors, so it acts more like pg_dump scripts

20 years agoAdd function (actually an int4 and an int8 version) that generates
Joe Conway [Tue, 3 Feb 2004 08:29:57 +0000 (08:29 +0000)]
Add function (actually an int4 and an int8 version) that generates
a series of numbers, optionally using an explicit step size other
than the default value (one). Use function in the information_schema
to replace hard-wired knowledge of INDEX_MAX_KEYS. initdb forced due
to pg_proc change. Documentation update still needed -- will be
committed separately.

20 years agoWin32 adjustments for dirmod and configure/configure.in.
Bruce Momjian [Mon, 2 Feb 2004 22:20:33 +0000 (22:20 +0000)]
Win32 adjustments for dirmod and configure/configure.in.

20 years agoautoconf run.
Tom Lane [Mon, 2 Feb 2004 18:35:45 +0000 (18:35 +0000)]
autoconf run.

20 years agoSet only LC_ALL for sort in tags creation.
Bruce Momjian [Mon, 2 Feb 2004 17:29:32 +0000 (17:29 +0000)]
Set only LC_ALL for sort in tags creation.